xavi/android_kernel_m2note/drivers/md/dm-crypt.c, branch o-8.1

dm-crypt: remove io_pool

2017-12-27T15:10:45+00:00

Remove io_pool and _crypt_io_pool because they are unused. CRs-fixed: 670391 Change-Id: I71400ecda66902c56c3af981e8b739f156db1e27 Signed-off-by: Mikulas Patocka Patch-mainline: dm-devel @ 04/05/14, 14:08 Signed-off-by: Joonwoo Park

dm-crypt: sort writes

2017-12-27T15:10:45+00:00

Write requests are sorted in a red-black tree structure and are submitted in the sorted order. In theory the sorting should be performed by the underlying disk scheduler, however, in practice the disk scheduler accepts and sorts only 128 requests. In order to sort more requests, we need to implement our own sorting. CRs-fixed: 670391 Change-Id: Iffd9345fa1253f5cf2a556893ed36e08f1ac51aa Signed-off-by: Mikulas Patocka Patch-mainline: dm-devel @ 04/05/14, 14:09 Signed-off-by: Joonwoo Park

dm-crypt: offload writes to thread

2017-12-27T15:10:45+00:00

Submitting write bios directly in the encryption thread caused serious performance degradation. On multiprocessor machine encryption requests finish in a different order than they were submitted in. Consequently, write requests would be submitted in a different order and it could cause severe performance degradation. This patch moves submitting write requests to a separate thread so that the requests can be sorted before submitting. Sorting is implemented in the next patch. Note: it is required that a previous patch "dm-crypt: don't allocate pages for a partial request." is applied before applying this patch. Without that, this patch could introduce a crash. CRs-fixed: 670391 Change-Id: I886ed2da0ff174d3539ea18e27170d7fd1062680 Signed-off-by: Mikulas Patocka Patch-mainline: dm-devel @ 04/05/14, 14:08 Signed-off-by: Joonwoo Park

dm-crypt: avoid deadlock in mempools

2017-12-27T15:10:45+00:00

This patch fixes a theoretical deadlock introduced in the previous patch. The function crypt_alloc_buffer may be called concurrently. If we allocate from the mempool concurrently, there is a possibility of deadlock. For example, if we have mempool of 256 pages, two processes, each wanting 256, pages allocate from the mempool concurrently, it may deadlock in a situation where both processes have allocated 128 pages and the mempool is exhausted. In order to avoid this scenarios, we allocate the pages under a mutex. In order to not degrade performance with excessive locking, we try non-blocking allocations without a mutex first and if it fails, we fallback to a blocking allocation with a mutex. CRs-fixed: 670391 Change-Id: I6c391dece4ba44fe0b2e9b75ea2b9235bf1b525b Signed-off-by: Mikulas Patocka Patch-mainline: dm-devel @ 04/05/14, 14:07 Signed-off-by: Joonwoo Park

dm-crypt: don't allocate pages for a partial request

2017-12-27T15:10:45+00:00

This patch changes crypt_alloc_buffer so that it always allocates pages for a full request. This change enables further simplification and removing of one refcounts in the next patches. Note: the next patch is needed to fix a theoretical deadlock CRs-fixed: 670391 Change-Id: I7bcadac8b3450976366c701fceb1fee7cb18df85 Signed-off-by: Mikulas Patocka [joonwoop@codeaurora.org: resolve trivial merge conflicts] Patch-mainline: dm-devel @ 04/05/14, 14:07 Signed-off-by: Joonwoo Park

dm-crypt: use per-bio data

2017-12-27T15:10:45+00:00

This patch changes dm-crypt so that it uses auxiliary data allocated with the bio. Dm-crypt requires two allocations per request - struct dm_crypt_io and struct ablkcipher_request (with other data appended to it). It used mempool for the allocation. Some requests may require more dm_crypt_ios and ablkcipher_requests, however most requests need just one of each of these two structures to complete. This patch changes it so that the first dm_crypt_io and ablkcipher_request and allocated with the bio (using target per_bio_data_size option). If the request needs additional values, they are allocated from the mempool. CRs-fixed: 670391 Change-Id: I8abc48a021391398f3b35bdd4ac9efbbec3a9fa5 Signed-off-by: Mikulas Patocka Patch-mainline: dm-devel @ 04/05/14, 14:05 Signed-off-by: Joonwoo Park

dm-crypt: run in a WQ_HIGHPRI workqueue

2017-12-27T14:45:03+00:00

Running dm-crypt in a standard workqueue results in IO competing for CPU time with standard user apps, which can lead to pipeline bubbles and seriously degraded performance. Move to a WQ_HIGHPRI workqueue to protect against that. bug 25392275 Change-Id: I589149a31c7b5d322fe2ed5b2476b1f6e3d5ee6f

dm-crypt: use unbound workqueue for request processing

2017-12-27T14:45:03+00:00

Use unbound workqueue so that work is automatically ballanced between available CPUs. CRs-fixed: 670391 Change-Id: I169099d0b5b27535633c9d3aaab2037b5fea6aa9 Signed-off-by: Mikulas Patocka [joonwoop@codeaurora.org: resolve trivial merge conflict] Patch-mainline: dm-devel @ 04/05/14, 14:06 Signed-off-by: Joonwoo Park

dm crypt: mark key as invalid until properly loaded

2017-07-02T17:38:49+00:00

commit 265e9098bac02bc5e36cda21fdbad34cb5b2f48d upstream. In crypt_set_key(), if a failure occurs while replacing the old key (e.g. tfm->setkey() fails) the key must not have DM_CRYPT_KEY_VALID flag set. Otherwise, the crypto layer would have an invalid key that still has DM_CRYPT_KEY_VALID flag set. Signed-off-by: Ondrej Kozina Reviewed-by: Mikulas Patocka Signed-off-by: Mike Snitzer Signed-off-by: Willy Tarreau

first commit

2016-08-15T02:19:42+00:00